Balancing Costs and Driver Satisfaction in Cargo Train Driver Scheduling

Abstract

This study considers the problem of scheduling cargo train drivers to perform a given set of activities, which is an important problem for our client Rail Force One. In this multi-objective problem, a trade-off must be found between monetary costs and driver satisfaction. Solutions must adhere to many real-world constraints, like maximum shift lengths, minimum resting times, contract hours, and driver qualifications. They must also deal with multiple types of drivers with different starting locations, varying salary rates throughout the time frame, and differing scheduling rules. To solve this problem, we developed a simulated annealing algorithm to find high-quality schedules. This algorithm includes a simplified but high-performance robustness model that focuses on the idle time between consecutive activities, as well as a satisfaction score comprised of twelve criteria for each driver. By using varying weights between estimated costs and satisfaction throughout the simulated annealing search, a Pareto-optimal front of schedules is created representing the trade-off between the two objectives. Experimental analysis using real-world data shows that the results of the algorithm are feasible and produced in acceptable runtimes. The generated schedules outperform the client's human-made schedules in cost-effectiveness, driver satisfaction and robustness. Consequently, the algorithm can reduce planning staff workload while increasing the client's profits and their drivers' satisfaction.